Inconsistent responses from matrix API when route not found

[rrrnld]

I am receiving a duration of null when requesting the time-distance matrix between these two points:

$  http https://api.openrouteservice.org/matrix profile==driving-car api_key==my-api-key range==1 locations=='14.101061,53.29817|12.808634,52.927325' sources==0

Response:

{
  "durations": [
    [
      0,
      null
    ]
  ],
  "destinations": [
    {
      "location": [
        14.101061,
        53.29817
      ],
      "snapped_distance": 0.03
    },
    {
      "location": [
        12.808634,
        52.927325
      ],
      "snapped_distance": 0
    }
  ],
  "sources": [
    {
      "location": [
        14.101061,
        53.29817
      ],
      "snapped_distance": 0.03
    }
  ],
  "info": {
    "service": "matrix",
    "engine": {
      "version": "4.5.0",
      "build_date": "2018-03-26T14:15:39Z"
    },
    "attribution": "openrouteservice.org | OpenStreetMap contributors",
    "timestamp": 1528366746044,
    "osm_file_md5_hash": "783b5a0ed086d6cc3eb6f98df9d5d79b",
    "query": {
      "profile": "driving-car",
      "units": "m"
    }
  }
}

How should I interpret this? I didn't find anything about it in the API docs.

[aoles]

null indicates no route was found, as documented:

The result is null if a value can't be determined.

A query to directions with the same coordinates returns

"error": {
    "code": 2009,
    "message": "Route could not be found - Unable to find a route between points 1 (14.1010610 53.2981700) and 2 (12.8086340 52.9273250)."
  }

so actually the result you are receiving from matrix is the expected one.

However, there seem to be some problem with the response because for sources=1 I get

"error": {
    "code": 6099,
    "message": "Unable to compute a distance/duration matrix."
  }

while sources=all yields

"durations": [
    [
      0,
      0
    ],
    [
      0,
      0
    ]
  ]

It' also puzzling why no route between the given coordinates can be found in first place, even though the locations seem to be legit.

[HendrikLeuschner]

I will have a look at it in the next couple of weeks.

[rabidllama]

@HendrikLeuschner any news on this yet?

[HendrikLeuschner]

I have built the Brandenburg graphs locally and run the routing query that give the route not found, but it works for me. The matrix query does also produce correct (the same as the routing) results. Need more investigation for why it does not work on the api. It might have something to do with the osm file...

[rabidllama]

@HendrikLeuschner Is this still an issue after the 4.7 update?

[HendrikLeuschner]

Will check out this week. Sorry for the delay @rabidllama

[HendrikLeuschner]

The problem is something with the second node at Karl-Marx-Strasse. A normal routing query https://maps.openrouteservice.org/directions?n1=52.927022&n2=12.809844&n3=17&a=52.926485,12.80758,53.29817,14.101061&b=0&c=0&k1=en-US&k2=km does not find the route either. Setting the point to a different location in Neuruppin works fine. It seems there are two nodes the iterator on the virtual node can go to (as expected). But both return infinite weighting, no matter if shortest or fastest. Will look further into this.

[HendrikLeuschner]

I think the problem is that the speed is calculated from an initial value of 5 down to 2.25. Then there is an if(speed < speedEncoder.factor/2) which is 2.25<2.5. Therefore it sets it to 0. That results in an edge that cannot be accessed.

Edit: Yeah, the residential penalty that is added drops the speed below the required minimum. Maybe we should apply the residential thing only when the actual speed is higher than lets say 10? @rabidllama

[HendrikLeuschner]

Fixed by #357

[rabidllama]

Will reopen until the code is in the master branch and released on the front end